{
  "dataset_name": "allenai/tulu-v2-sft-mixture",
  "learning_rate": 0.0001,
  "lr_scheduler_type": "linear",
  "warmup_ratio": 0.03,
  "weight_decay": 0.0,
  "num_train_epochs": 5,
  "per_device_train_batch_size": 1,
  "gradient_accumulation_steps": 16,
  "max_seq_length": 4096
}